Home » News » When Rules are Meant to Be Broken

When Rules are Meant to Be Broken


May 15, 2024 | Originally published on Empire Startups Fintech Newsletter

Throughout history, we have seen if it wasn’t for rule breakers, many pivotal aspects of our lives would not be as they are today. Some would call these rule breakers innovators, others disruptors. The difference between innovators and disruptors ultimately lies in a deep understanding of the distinction between what rules must be followed diligently as opposed to the rules that can (and should) be rewritten. 

In the realm of finance, where stability and predictability are crucial, rule-breaking takes on a nuanced significance. While disruptive innovation can introduce instability, conscientious innovation and disciplined monitoring hold the potential to redefine our financial landscape for the better. 

As the youngest child in my family, I was a classic example of a people pleaser in my early years. I followed the rules to get praise and some amount of attention. Then, at some point in my teenage years, I started to recognize where I could bend the rules without breaking them, but just enough to usher in incremental opportunity (and fun). In the two decades I’ve spent in finance and the work we’re doing at Stratyfy, knowing this distinction has been critical in driving the right kind of change, change that drives the industry forward while understanding the unique situation (and rules) in which our financial system (and our customers) must operate within. 

To say the way lending decisions are being made is broken would be a discredit to the many years of work invested in finding the ideal balance between growth and risk management necessary to make sound lending decisions. Yet, in those years, lenders have stuck to credit policies or rules that only benefit a certain portion of consumers. With these rules providing a comfortable and familiar safety net for risks, it is understandable why it would be much easier to continue relying on the same methods to do business.

But what happens when borrower profiles change faster than that of the methodologies being used to evaluate and make decisions about those borrowers? 

Fellow contributor Craig J. Lewis highlighted in his piece last fall, Trust: The Currency of the Gig Economy, that in the US alone there were an estimated 73.3 million freelancers, or 43% of the total labor force, in 2023.  A structural shift towards an independent type of work and income.  Another contributor, Ron J Williamsreminded us earlier this year how small businesses have always been the backbone of the American economy. What is happening to this growing number of individuals deciding to take the plunge and work for themselves? More advanced methods for making risk based decisions are a must as lenders have struggled to adequately support these businesses with responsible and efficient lending products.

Borrowers have clearly changed, but the way in which the majority of lenders make loan decisions has not.

In the example above, based on Capacity within the 5 C’s of credit, the freelancer’s varying income would likely lower their ability to get approved for a loan. So many loan decisions, particularly rejections, are made based on sharp cut-offs that fail to capture the nuances associated with the ways borrowers have changed over the past decade.

This is where AI, more specifically interpretable machine learning, has the power to unlock value for both lenders and borrowers.  However, challenging ingrained beliefs and asking people to do the hardest thing –change – is always going to come with a bit of push back.

Take for example, how lending decisions are made. There has been a lot of talk of the value AI, and more specifically Machine Learning (“ML”), can bring to make lending decisions more profitable, efficient and fair. But most of this value has not been realized, particularly by community and regional banks, in part because many ML providers approached these lenders with solutions that throw away the old in favor of the new. They do not meet the lender where they are, offering trustworthy and reliable improvements over time, and instead try to break too many rules. This is not the recommended approach for inspiring change that is also safe – breaking some of the rules, but not all of them at the same time. 

The other reason that ML has struggled to deliver the value promised is the fact that finance, and lending in particular, is often slow to move but fast to follow when something has been proven by the masses. This gives Financial institutions (“FIs”) the confidence, strength in numbers and social validation that is often necessary to break away from the rules of yesterday. While most FIs historically have learned to get better through “interpolation,” leading FI innovators have achieved success with “extrapolation” – taking calculated risks in developing products and decision strategies outside of their historical data and policies (aka rules).

The goal of the AI should be to augment human intelligence, not replace it, with the best solutions having the ability to learn from both data AND subject matter expertise.

“AI meets IQ”.

Beyond this overall philosophy of human + machine delivering superior results, there are specific attributes that financial institutions should prioritize when evaluating AI technology to ensure the chosen technology fits the business problem and regulatory environment at hand.

The most important ones for high stakes decisions like lending are:

(1) predictive power, in the field not just the lab, 

(2) robustness, 

(3) transparency, recognizing that all types of transparency are not created equally and 

(4) fairness, both on an individual and group basis. 

In the context of lending, #3 is particularly essential. And the right kind of transparency, interpretability, has the ability to also help drive results in 1, 2 and 4 (in addition to 3, obvi). This prioritization will allow lenders to select the right technology that follows the rules it needs to follow and breaks away from the rules that are holding lenders (and their borrowers) back.

With the only constant being change, borrowers will continue to evolve with both newer and older generations having very different financial profiles than the past. ML technology, if selected, developed and deployed properly, is positioned to completely reinvent the way lenders evaluate creditworthiness and make decisions on loans, ensuring these lenders can unlock profitable, inclusive growth and deliver for the borrowers of today and tomorrow.